SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillslovstudio-rich-export

lovstudio-rich-export

Office & productivity

View repository

Updated 10-16 00:01

About this skill

One-Click Rich Media Export

First converge the content into an auditable source, then generate for each delivery format a version it can faithfully express. HTML preserves interactive media; DOCX, PDF, and Markdown generate editable or printable static projections, and preserve covers, descriptions, and original links for media.

Triggers

  • The user requests one-click export of a piece of content or product data as HTML, Markdown, DOCX, PDF, ZIP, or an offline delivery package.
  • The user requests exporting content containing images, audio, video, iframes, interactive charts, or media descriptions, and needs correct handling for different formats.
  • The user needs to integrate reusable multi-format export capability into a web, desktop, or server-side product.

Do not activate when

  • Only need to edit a local part of an existing DOCX, PDF, or web page, and multi-format delivery is not involved.
  • Only need to generate a single image, a single slide, or a single video file; use the corresponding media production workflow.

First Define Content Boundaries

  1. Treat names, project background, internal status, prompts, and the creative process in the chat as internal context; only fields that the user requests to appear in the final product enter the input content.
  2. When generating titles, authors, covers, footers, and file names, only read explicit document metadata; do not infer or add personal information from the chat background.
  3. When there is video, audio, a map, an embedded web page, or an interactive chart, first declare its static projection. When there is no cover, PDF/DOCX should at least preserve the name, summary, and accessible link.

Input and Commands

The default input is .md; .html and the rich-export.jsondefined by this Skill are also supported. Relative resources in the input are resolved relative to the input file.

python3 "$SKILL_DIR/scripts/export_rich.py" \
  --input ./release/article.md \
  --out ./release/exports \
  --formats html-single,html-dir,md,docx,pdf \
  --zip

First use --formats to generate only the format the user needs.html-single is suitable for direct sending and offline opening;html-dir is suitable when videos are large, there are many files, or the original media files need to be preserved;docx is suitable for continued editing;pdf is suitable for printing and archiving;md is the traceable source document.

Run python3 "$SKILL_DIR/scripts/export_rich.py" --help to view dependencies and options. For the complete input contract, see references/format-contract.md, and for product integration, see references/product-integration.md。

Rendering Strategy

Target Generation Method Media Behavior
Single-file HTML Pandoc --embed-resources Embed images, styles, scripts, audio, and video; large files will expand significantly
HTML folder Pandoc + assets/ Preserve interactivity and independent resources, suitable for actual distribution
Markdown Pandoc normalization Preserve images and links; media uses explicit links
DOCX Pandoc + optional reference.docx Images are embedded; video/audio/iframe are projected as covers, descriptions, and links
PDF Chromium/Playwright prints static HTML Images and layout are preserved; media is projected as covers, descriptions, and links

Do not treat single-file HTML as the default delivery for large videos. When the total media size exceeds about 20 MB, by default also deliver html-dir; if the user must receive only one file, deliver a ZIP containing the HTML folder and export-manifest.json。

Acceptance

Perform the following checks after each export:

  1. Open the single-file HTML and refresh with the network disconnected; check fonts, images, audio/video controls, table of contents, and links.
  2. Unzip/open the HTML folder; check that all resources of index.html come from within the package and that there are no absolute local paths.
  3. Open the DOCX with Word/LibreOffice; check Chinese fonts, tables, images, pagination, and every media link.
  4. Render the PDF's first page, pages containing images, pages containing media projections, and the last page; check for cropping, overflow, garbled text, links, and page numbers.
  5. Check the export-manifest.json of warnings; internal metadata or prompts must not be exported as body text.

Choices and Limitations

  • By default, Pandoc + Playwright is used: the former handles structured format conversion, and the latter reliably prints the confirmed web layout to PDF.
  • The existing lovstudio-any2pdf、lovstudio-any2docx is still suitable for specialized artistic layout of long Chinese text; this Skill handles multi-format single-source delivery and rich media fallback strategies.
  • SingleFile CLI is not made a built-in product dependency: its open-source repository is AGPL, and it is better suited to web archiving, not a multi-format content publishing pipeline.
  • HTML and print/office formats are not equivalent. Interactive charts, iframes, and media playback capabilities are only preserved faithfully in HTML; other formats should deliver the information and access paths completely, rather than pretending to be playable content.